- Email: [email protected]
A note on blocks with abelian defect groups
Journal of Algebra 317 (2007) 250–259 www.elsevier.com/locate/jalgebra
A note on blocks with abelian defect groups Yun Fan a , Burkhard Külshammer b,∗ a Department of Mathematics, Central China Normal University, Wuhan 430079, China b Mathematical Institute, Friedrich Schiller University, 07737 Jena, Germany
Received 29 October 2006 Available online 13 April 2007 Communicated by Michel Broué Dedicated to Michel Broué on the occasion of his 60th birthday
Abstract A recent result by H. Meyer shows that, for a field F of characteristic p > 0 and a finite group G with an abelian Sylow p-subgroup, the F -subspace Zp F G of the group algebra F G spanned by all p-regular class sums in G is multiplicatively closed, i.e. a subalgebra of the center ZF G of F G. Here we generalize this result to blocks. More precisely, we show that, for a block A of a group algebra F G with an abelian defect group, the F -subspace Zp A := A ∩ Zp F G is multiplicatively closed, i.e. a subalgebra of the center ZA of A. We also show that this subalgebra is invariant under perfect isometries and hence under derived equivalences. © 2007 Elsevier Inc. All rights reserved. Keywords: Representation theory of finite groups; Block theory
1. Introduction In a recent paper [8], H. Meyer proved that, for a field F of characteristic p > 0 and a finite group G with an abelian Sylow p-subgroup P , the F -subspace Zp F G of the group algebra F G spanned by all p-regular class sums in G is multiplicatively closed, i.e. a subalgebra of the center ZF G of F G. His result was motivated by an earlier paper by J. Murray [9] who showed that Zp F G is multiplicatively closed for G a symmetric or alternating group. And Murray’s * Corresponding author.
E-mail address: [email protected] (B. Külshammer). 0021-8693/$ – see front matter © 2007 Elsevier Inc. All rights reserved. doi:10.1016/j.jalgebra.2007.03.043
Y. Fan, B. Külshammer / Journal of Algebra 317 (2007) 250–259
251
proof made use of results in [4] and [5]. On the other hand, H. Meyer gives several examples of finite groups G such that Zp F G is not multiplicatively closed. In this paper, we will consider the F -subspace Zp A := A ∩ Zp F G of the center ZA of A, for a block A of a group algebra F G. As is well known, the dimension of Zp A coincides with the number of simple A-modules. Our main result will be a generalization of Meyer’s theorem to blocks. In Section 2 we show that, for a block A with an abelian defect group D (in a finite group with an arbitrary Sylow p-subgroup P ), the F -subspace Zp A is multiplicatively closed, i.e. a subalgebra of ZA. This generalization is not quite straightforward since Meyer’s proof uses a transfer argument which is not directly applicable to blocks. Our way around this problem is to use results by Watanabe [10], Fan [3] and Külshammer–Okuyama–Watanabe [7] instead. Our main result has applications to perfect isometries, as defined by M. Broué [1]. In Section 3 we show that, for blocks A and B with abelian defect groups in finite groups G and H , respectively, the isomorphism of F -algebras ZB → ZA induced by a perfect isometry between A and B maps Zp B onto Zp A. In particular, this applies to all situations where Broué’s abelian defect group conjecture is known to hold. Hence the F -algebra Zp A provides a new invariant of perfect isometries and hence of derived equivalences, at least for blocks with abelian defect groups. 2. Abelian defect groups In the following, we fix a prime number p and a p-modular system (K, O, F ), that is, O is a complete discrete valuation ring with field of fractions K of characteristic 0 and residue field F of characteristic p. The O-algebras we consider will always be free of finite rank as O-modules. It will be important, at a key step for our main result, to work over O, and not just over F , though the following first two results hold not only for O. We start with a somewhat technical result on abstract algebras whose relevance will become clear later. 2.1. Proposition. Let A be an O-algebra, and let L be a split local unitary subalgebra of the center ZA of A. Moreover, let e be an idempotent in A such that AeA = A, and suppose that C is a unitary subalgebra of eAe such that the multiplication map μC : L ⊗O C → eAe,
z ⊗ c → zc,
is an isomorphism of O-algebras. Then there exists a unitary subalgebra B of A such that C = eBe and such that the multiplication map μB : L ⊗O B → A,
z ⊗ b → zb,
is an isomorphism of O-algebras. r Proof. We write e = lr=1 m s=1 frs with pairwise orthogonal primitive idempotents frs in C where frs and fr s are conjugate in C if and only if r = r . Since L is local and since μC is an isomorphism, every frs is also primitive in eAe and thus in A. Moreover, frs and fr s are conjugate in A if and only if they are conjugate in eAe if and only if they are conjugate in C. Next, note that every primitive idempotent of A is conjugate to one in eAe since AeA = A; so we combine our decomposition of e with a similar decomposition of 1 − e in order to obtain
252
Y. Fan, B. Külshammer / Journal of Algebra 317 (2007) 250–259
r a decomposition 1 = lr=1 ns=1 frs into pairwise orthogonal primitive idempotents frs in A where frs and fr s are conjugate in A if and only if r = r , and nr mr for r = 1, . . . , l. We conclude that nr nr l
A=
frs Afr s ,
r,r =1 s=1 s =1
and eAe =
mr mr l r,r =1
s=1
frs Af
r s
,
C=
s =1
mr mr l r,r =1
s=1
frs Cfr s .
s =1
For r = 1, . . . , l and s = 1, . . . , nr , we can choose elements xrs ∈ frs Afr1 and yrs ∈ fr1 Afrs such that frs = xrs yrs and fr1 = yrs xrs ; here we may and will assume that xrs ∈ frs Cfr1 and yrs ∈ fr1 Cfrs whenever s mr . Then frs Afr s = xrs yrs Axr s yr s ⊆ xrs Ayr s ⊆ frs Afr s , and xrs Ayr s = xrs fr1 Afr 1 yr s ⊆ xrs (eAe)yr s ⊆ xrs Ayr s ; therefore, frs Afr s = xrs Ayr s = xrs (eAe)yr s for r, r = 1, . . . , l, s = 1, . . . , nr and s = 1, . . . , nr . And similarly, frs Cfr s = xrs Cyr s whenever s mr and s mr . This implies that, for r, r = 1, . . . , l, s = 1, . . . , nr , s = 1, . . . , nr , the maps xrs Ayr s → xr1 Ayr 1 ,
a → yrs axr s ,
xr1 Ayr 1 → xrs Ayr s ,
a → xrs ayr s ,
and
are mutually inverse isomorphisms of L-modules, which induce O-module isomorphisms xrs Cyr s ∼ = xr1 Cyr 1 . We claim that B :=
nr nr l
xrs Cyr s
r,r =1 s=1 s =1
is a unitary subalgebra of A satisfying the conclusion of the proposition. In fact, we have 1=
nr l r=1 s=1
frs =
nr l r=1 s=1
xrs fr1 yrs ∈
nr l
xrs Cyrs ⊆ B.
r=1 s=1
Moreover, we see that xrs Cyr s · xr s Cyr s = 0 unless r = r and s = s , and that xrs Cyr s · xr s Cyr s = xrs Cfr 1 Cyr s ⊆ xrs Cyr s ⊆ B
Y. Fan, B. Külshammer / Journal of Algebra 317 (2007) 250–259
253
for all possible r, r , r , r , s, s , s , s . Thus B is a unitary subalgebra of A containing C; hence C = eCe ⊆ eBe. On the other hand, we have exrs Cyr s e = 0 whenever s > mr or s > mr , and exrs Cyr s e = xrs Cyr s whenever s mr and s mr . This shows that eBe ⊆ C, and we have proved that eBe = C. At last, since the multiplication map μB is certainly a homomorphism of O-algebras, it is enough to show that the following restriction of μB is an O-isomorphism μB |L⊗O (xrs Cyr s ) : L ⊗O (xrs Cyr s ) → xrs (eAe)yr s ,
z ⊗ xrs cyr s → xrs zcyr s .
The hypothesis that the multiplication map μC is surjective implies that the above map is surjective; and it is injective because the following isomorphism implies that the O-ranks of both sides are equal: μC
L ⊗O (xrs Cyr s ) ∼ = L ⊗O (xr1 Cyr 1 ) ∼ = xr1 (eAe)yr 1 = fr1 Afr 1 ∼ = frs Afr s .
2
We can use Proposition 2.1 in order to prove the following structure theorem for certain blocks with abelian defect groups. 2.2. Theorem. Let G be a finite group, and let A be a block of the group algebra OG with maximal A-subpair (D, bD ). Suppose that D is abelian and that Q := CD (NG (D, bD )) ⊆ Z(G). Then there exists a unitary subalgebra B of A such that the multiplication map μ : OQ ⊗O B → A,
x ⊗ y → xy,
is an isomorphism of O-algebras. Proof. We fix an idempotent i in A such that iAi is a source algebra of A. Results of Fan [3] (which is proved without restrictions on the size of O) and Külshammer–Okuyama–Watanabe [7] imply that there exists a unitary subalgebra C of iAi such that the multiplication map OQ ⊗O C → iAi,
x ⊗ y → xy,
is an isomorphism of O-algebras. Since AiA = A, the result follows from Proposition 2.1.
2
In fact, the theorem above (for sufficiently large O) was already contained in a preliminary version of [7]; however, it was not incorporated into the final version. We also note that the O of A in O[G/Q], which is a block with algebra B above is clearly isomorphic to the image A defect group D/Q when O is large enough. In the following, we fix a finite group G and a block A = OGe of the group algebra OG, with block idempotent e and defect group D. We denote by Gp the set of p-regular elements in G, and by OGp the O-sublattice of OG spanned by Gp . In general, neither OGp nor Zp OG := ZOG ∩ OGp are multiplicatively closed (not even when the Sylow p-subgroups of G are abelian), as easy examples show. Note that the p-regular class sums form an O-basis of Zp OG. We set Zp A := A ∩ Zp OG = (Zp OG)e; the last equality follows from a result of Iizuka (cf. [6]).
254
Y. Fan, B. Külshammer / Journal of Algebra 317 (2007) 250–259
2.3. Corollary. In the situation of Theorem 2.2, we have Zp A ⊆ ZB. Proof. Let C be a p-regular conjugacy class of G, with class sum C + := c∈C c in OG. We need to show that C + e ∈ ZB. Since the multiplication map μ : OQ⊗O B → A is an isomorphism of O-algebras, we have decompositions A=
uB
and ZA = μ Z(OQ ⊗O B) = μ(OQ ⊗O ZB) = uZB.
u∈Q
u∈Q
First we assume that the group algebra O c for c ∈ C is split semisimple, i.e. is isomorphic to a direct product of copies of O; in particular, c is an O-linear combination of idempotents in O c . Then we can write C + e = nm=1 αm jm with idempotents jm ∈ A (not necessarily orthogonal) and coefficients αm ∈ O. Since A/J (A) ∼ = B/J (B), every idempotent of A is conjugate to one in B; so we may choose units um in A such that im := um jm u−1 m ∈ B, for m = 1, . . . , n. Then C+e =
n m=1
αm jm ≡
n
αm im =: b
mod [A, A] ,
m=1
where b ∈ B, and [A, A] is the O-sublattice of A spanned by all commutators xy −yx (x, y ∈ A). And note that [A, A] = μ [OQ ⊗O B, OQ ⊗O B] = μ OQ ⊗O [B, B] = u[B, B]. u∈Q
On the other hand, we may write C + e = u∈Q uzu with uniquely determined elements zu ∈ ZB. Then uzu − b = C + e − b ∈ [A, A] = u[B, B], u∈Q
u∈Q
and we conclude that zu ∈ [B, B] whenever 1 = u ∈ Q. Hence zu ∈ ZB ∩ [B, B] = 0 (since we are working in characteristic 0) whenever 1 = u ∈ Q, so we obtain C + e = z1 ∈ ZB. Next assume that O c for c ∈ C is not split. Since c is p-regular, there is a finite extension O is split semisimple; hence, in of O such that O c ⊗O A = := O A
u∈Q
⊗O B) = u(O
uB
u∈Q
⊗O B, we have C + e ∈ ZB := O (the arguments of the previous paragraph are in fact where B based on the structure A = u∈Q uB and independent of the two-sided indecomposability of A). So
= = ZB. uZB ∩ ZB 2 C + e ∈ ZA ∩ ZB u∈Q
Now we turn our attention to characteristic p. In the following, we denote the canonical maps O → F and OG → F G by x → x. Moreover, we denote by F Gp the F -subspace of the group
Y. Fan, B. Külshammer / Journal of Algebra 317 (2007) 250–259
255
algebra F G spanned by Gp . Then Zp F G := ZF G ∩ F Gp is the F -subspace of ZF G spanned by the p-regular class sums of G. Since A is a block of OG with defect group D, its image A is a block of F G with defect group D. We are interested in the F -subspace Zp A := A ∩ Zp F G = (Zp F G)e of ZA. Note that Zp F G = Zp OG and Zp A = Zp A. We call A p -closed if Zp A is multiplicatively closed, i.e. a subalgebra of ZA. Similarly, we call OG p -closed if Zp F G is a subalgebra of ZF G. By H. Meyer’s result [8], group algebras of groups with abelian Sylow p-subgroups are p -closed. We prove the following generalization. 2.4. Theorem. Blocks with abelian defect groups are p -closed. Proof. First we assume that O is large enough. Let A = OGe be a minimal counterexample, where e denotes the block idempotent of A. Let (D, bD ) be a maximal A-subpair, so that D is a defect group of A. Since D is abelian, we have D = Q × R where Q := CD (NG (D, bD )) and R := [D, NG (D, bD )]. Assume that z1 , z2 ∈ Zp A but z1 z2 ∈ / Zp A, and let z1 z2 =
zg g,
zg ∈ F.
g∈G
Then there is a p-singular element g ∈ G such that zg = 0. Writing g = su = us with a p-element u ∈ G and a p-regular element s ∈ CG (u), and applying the Brauer homomorphism Br u , we obtain Br u (z1 ), Br u (z2 ) ∈ Zp F CG (u)Br u (e), but Br u (z1 )Br u (z2 ) ∈ / Zp F CG (u)Br u (e). Since the defect groups of the blocks appearing in F CG (u)Br u (e) ¯ are still abelian, one of these blocks is still a counterexample to the theorem. By the minimality of the counterexample A = OGe, it must be the case that CG (u) = G. So u belongs to D and is centralized by NG (D, bD ), hence belongs to Q. We conclude that the p-factor of any g ∈ G such that zg = 0 belongs to Q. Let bQ be the unique block of OCG (Q) such that (Q, bQ ) (D, bD ). Then a result by Watanabe [10] implies that the map β : ZA → ZbQ ,
y → BrQ (y)1bQ ,
is an isomorphism of F -algebras; this isomorphism maps Zp A onto Zp bQ , so that bQ is also a counterexample to the theorem. Thus it must be the case that CG (Q) = G, in other words, Q is a central p-subgroup of G. Now, on one hand we may write z1 z2 =
u∈Q
us u =
u∈Q
us u e
256
Y. Fan, B. Külshammer / Journal of Algebra 317 (2007) 250–259
with s u ∈ Zp F G, hence s u e ∈ Zp A for u ∈ Q. On the other hand, by Theorem 2.2, we have a unitary subalgebra B of A such that A= uB, and ZA = uZB, u∈Q
u∈Q
where B denotes the image of B in A. Then, by Corollary 2.3, we have Zp A ⊆ ZB and therefore Zp A ⊆ ZB. In particular, ZB contains z1 , z2 and s u e for u ∈ Q. But then also z1 z2 ∈ ZB. Since ZA = u∈Q uZB we conclude that s u e = 0 whenever 1 = u ∈ Q. This shows that z1 z2 = s 1 e ∈ Zp A, and we have reached a contradiction. and extend A to If O is not large enough, then we can extend it to a large enough one O, ⊗O A. It is easy to see that = O A = Zp A. ZA ∩ Zp A decompose e as a sum of primitive central idempotents: e = e1 + · · · + en ; then e1 , . . . , en In A, t = Ae t for t = 1, . . . , n are blocks with defect are conjugate under a suitable Galois group, and A group D. For z1 , z2 ∈ Zp A, by the conclusion proved above, t , z1 z2 et ∈ Zp A
t = 1, . . . , n;
so z1 z2 =
n t=1
z1 z2 e t ∈
n
t = Zp A; Zp A
t=1
in conclusion, we get = Zp A. z1 z2 ∈ ZA ∩ Zp A
2
3. Perfect isometries In this section we assume that K and F are splitting fields for all the algebras we will consider, and we point out a connection between Theorem 2.4 and perfect isometries, as defined by M. Broué [1]. We start by recalling the definition of a perfect isometry and some of its consequences. Suppose that A = OGe and B = OHf are blocks of finite groups G and H , respectively; here e and f are the corresponding block idempotents. We denote the set of irreducible characters of G in A by Irr(A), the group of virtual characters of G in A by Z Irr(A), and the set of class functions G → K in A by CF(A, K). An isometry I : Z Irr(B) → Z Irr(A) (with respect to the canonical inner products) is called perfect if the virtual character μ of G × H defined by (Iβ)(g) · β h−1 (g ∈ G, h ∈ H ), μ(g, h) := β∈Irr(B)
satisfies the following two conditions, for g ∈ G and h ∈ H :
Y. Fan, B. Külshammer / Journal of Algebra 317 (2007) 250–259
257
(Int) μ(g, h)/|CG (g)| ∈ O and μ(g, h)/|CH (h)| ∈ O. (Sep) If μ(g, h) = 0 then g and h are either both p-regular or both p-singular. Since I is an isometry, each β ∈ Irr(B) defines a sign β ∈ {1, −1} and an irreducible character αβ ∈ Irr(A) such that Iβ = β αβ . The map β → αβ is a defect-preserving bijection between Irr(B) and Irr(A), and I extends to a K-linear bijection, also denoted by I , between CF(B, K) and CF(A, K). Note that (I ψ)(g) = |H |−1
μ(g, h)ψ(h)
ψ ∈ CF(B, K), g ∈ G .
h∈H
Recall that the group algebra KG is a symmetric K-algebra with respect to the canonical bilinear form(· | ·) mapping a pair of elements x, y ∈ KG to the coefficient z1 ∈ K in the product xy = z = g∈G zg g. The restriction of (· | ·) turns KA := K ⊗O A into a symmetric K-algebra. Hence the vector spaces ZKA and CF(A, K) are isomorphic via the map associating to each class function φ ∈ CF(A, K) the element φ ◦ ∈ ZKA satisfying (φ ◦ |a) = φ(a) for a ∈ KA. Explicitly, we have φ g −1 g. φ◦ = g∈G
In a similar way, ZKB and CF(B, K) are in canonical K-linear bijection. Thus the K-linear bijection I : CF(B, K) → CF(A, K) yields a K-linear bijection I ◦ : ZKB → ZKA such that I ◦ ψ ◦ = (I ψ)◦ for ψ ∈ CF(B, K). For β ∈ Irr(B), let fβ = (β(1)/|H |) h∈H β(h−1 )h = (β(1)/|H |)β ◦ denote the primitive idempotent in ZKB corresponding to β. Similarly, for α ∈ Irr(A), let −1 α g g = α(1)/|G| α ◦ eα = α(1)/|G| g∈G
denote the primitive idempotent of ZKA corresponding to α. Then I ◦ fβ = β(1)/|H | I ◦ β ◦ = β(1)/|H | β αβ◦ = β β(1)/|H | |G|/αβ (1) eαβ for β ∈ Irr(B). Since f =
β∈Irr(B) fβ
I ◦f =
we conclude β β(1)/|H | |G|/αβ (1) eαβ .
β∈Irr(B)
In general, I ◦ is K-linear but not necessarily a homomorphism of K-algebras. It is easy to see that |H |−1 I ◦z = μ g −1 , h zh−1 g for z = zh h ∈ ZKB. g∈G
h∈H
h∈H
Thus the condition (Int) implies that I ◦ restricts to an O-linear map ZB → ZA which we will also denote by I ◦ .
258
Y. Fan, B. Külshammer / Journal of Algebra 317 (2007) 250–259
3.1. Lemma. In the situation above, we have I ◦ (Zp B) = Zp A. / Hp , and I ◦ z = g∈G yg g with Proof. Let z = h∈H zh h ∈ Zp B. Then zh = 0 whenever h ∈ yg = |H |−1 h∈H μ(g −1 , h)zh−1 for g ∈ G. If yg = 0 then 0 = μ(g −1 , h)zh−1 for some h ∈ H , so h ∈ Hp since zh−1 = 0. But now (Sep) implies that g ∈ Gp , and we have proved that I ◦ (Zp B) ⊆ Zp A. Thus the lemma follows by symmetry. 2 ◦
The O-linear bijection I ◦ : ZB → ZA induces an F -linear bijection I : ZB → ZA such that ◦ I (Zp B) = Zp A. In general, I is not a homomorphism of F -algebras, as easy examples show. On the other hand, the perfect isometry I between A and B induces an isomorphism of K-algebras ◦
ι : ZKB → ZKA in such a way that ι(fβ ) = eαβ for β ∈ Irr(B). The maps ι and I ◦ are related by the formula (see [1]) ι(z) = I ◦ zR ◦ (e) for z ∈ ZKB; here R = I −1 denotes the perfect isometry inverse to I . Hence ι restricts to an isomorphism of O-algebras ZB → ZA also denoted by ι. Note that e ∈ Zp A by a result of Osima (cf. [6]). Thus R ◦ (e) ∈ Zp B by Lemma 3.1. We denote by ι : ZB → ZA the isomorphism of F -algebras ◦ ◦ induced by ι : ZB → ZA. Then ι(z) = I (zR (e)) for z ∈ ZB. 3.2. Proposition. Let I be a perfect isometry between the blocks A = OGe and B = OHf , and suppose that B is p -closed. Then A is also p -closed, and the isomorphism of F -algebras ι : ZB → ZA defined by I satisfies ι(Zp B) = Zp A. ◦
◦
Proof. Let z ∈ Zp B. Since R (e) ∈ Zp B and since B is p -closed, we conclude that zR (e) ∈ ◦ ◦ Zp B. Thus Lemma 3.1 implies that ι(z) = I (zR (e)) ∈ Zp A. So we have proved that ι(Zp B) ⊆ Zp A. Since dim Zp B = dim Zp A, we obtain ι(Zp B) = Zp A, and the second assertion is proved. Since Zp A is the image of the subalgebra Zp B under the isomorphism of F -algebras ι, it is certainly a subalgebra of ZA, and the result follows. 2 We obtain the following consequence. 3.3. Corollary. Let I be a perfect isometry between blocks A = OGe and B = OHf of finite groups G and H , respectively, and suppose that the defect groups of B are abelian. Then A and B are p -closed, and the isomorphism of F -algebras ι : ZB → ZA defined by I maps Zp B onto Zp A. Proof. By Theorem 2.4, the block B is p -closed. Hence the result follows from Proposition 3.2. 2 The result above implies that the F -algebra Zp A is an invariant of perfect isometries and hence an invariant of derived equivalences, at least for blocks with abelian defect groups. (We
Y. Fan, B. Külshammer / Journal of Algebra 317 (2007) 250–259
259
recall that the dimension of Zp A is just the number of simple A-modules.) In the situation of Corollary 3.3, it is not true, however, in general, that ι(Zp B) = Zp A, as elementary examples show. Of course, one would expect that, in the situation of Corollary 3.3, the defect groups of A are also abelian. (This would follow, for example, from Brauer’s height zero conjecture.) One would perhaps even expect that A and B have isomorphic defect groups. The result above motivates the following problem. 3.4. Question. Let ι : ZB → ZA be the isomorphism of F -algebras defined by a perfect isometry I between blocks A = OGe and B = OHf of finite groups G and H , respectively. Is ι(Zp B) = Zp A? J. Murray [9] has proved that blocks of finite symmetric and alternating groups are p -closed. Thus, by Proposition 3.2, Question 3.4 has a positive answer whenever H is a symmetric or alternating group. (We also note that M. Enguehard has proved that p-blocks of the same weight in finite symmetric groups are always perfectly isometric [2].) One may view Question 3.4 as a problem concerning the p-sections of 1 in G and H , respectively. Thus one may ask whether similar properties hold for other p-sections as well, at least in the presence of an isotypy (cf. [1]). However, this does not seem to be the case, as easy examples show, not even for blocks with abelian defect groups in the situation of Broué’s abelian defect group conjecture. Acknowledgments The results in this paper were obtained during visits of the first author in Jena and of the second author in Wuhan. Both visits were made possible by a Sino-German exchange program dedicated to the theory of finite groups and their representations, funded by the NSFC and the DFG. Both authors gratefully acknowledge this support. The second author also profited from several discussions with G. Cliff, on the occasion of a visit to the University of Alberta in Edmonton and on the occasion of a visit by G. Cliff in Jena. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10]
M. Broué, Isométries parfaites, types de blocs, catégories dérivées, Astérisque 181–182 (1990) 61–92. M. Enguehard, Isométries parfaites entre blocs de groupes symétriques, Astérisque 181–182 (1990) 157–171. Y. Fan, Relative local control and the block source algebras, Sci. China Ser. A 40 (1997) 785–798. B. Külshammer, Bemerkungen über die Gruppenalgebra als symmetrische Algebra II, J. Algebra 75 (1982) 59–69. B. Külshammer, Group-theoretical descriptions of ring-theoretical invariants of group algebras, in: Progr. Math., vol. 95, 1991, pp. 425–442. B. Külshammer, Central idempotents in p-adic group rings, J. Aust. Math. Soc. Ser. A 56 (1994) 278–289. B. Külshammer, T. Okuyama, A. Watanabe, A lifting theorem with applications to blocks and source algebras, J. Algebra 232 (2000) 299–309. H. Meyer, On a subalgebra of the centre of a group ring, J. Algebra 295 (2006) 293–302. J. Murray, Squares in the centre of the group algebra of a symmetric group, Bull. London Math. Soc. 34 (2002) 155–164. A. Watanabe, Note on a p-block of a finite group with abelian defect group, Osaka J. Math. 26 (1989) 829–836.
A note on blocks with abelian defect groups Yun Fan a , Burkhard Külshammer b,∗ a Department of Mathematics, Central China Normal University, Wuhan 430079, China b Mathematical Institute, Friedrich Schiller University, 07737 Jena, Germany
Received 29 October 2006 Available online 13 April 2007 Communicated by Michel Broué Dedicated to Michel Broué on the occasion of his 60th birthday
Abstract A recent result by H. Meyer shows that, for a field F of characteristic p > 0 and a finite group G with an abelian Sylow p-subgroup, the F -subspace Zp F G of the group algebra F G spanned by all p-regular class sums in G is multiplicatively closed, i.e. a subalgebra of the center ZF G of F G. Here we generalize this result to blocks. More precisely, we show that, for a block A of a group algebra F G with an abelian defect group, the F -subspace Zp A := A ∩ Zp F G is multiplicatively closed, i.e. a subalgebra of the center ZA of A. We also show that this subalgebra is invariant under perfect isometries and hence under derived equivalences. © 2007 Elsevier Inc. All rights reserved. Keywords: Representation theory of finite groups; Block theory
1. Introduction In a recent paper [8], H. Meyer proved that, for a field F of characteristic p > 0 and a finite group G with an abelian Sylow p-subgroup P , the F -subspace Zp F G of the group algebra F G spanned by all p-regular class sums in G is multiplicatively closed, i.e. a subalgebra of the center ZF G of F G. His result was motivated by an earlier paper by J. Murray [9] who showed that Zp F G is multiplicatively closed for G a symmetric or alternating group. And Murray’s * Corresponding author.
E-mail address: [email protected] (B. Külshammer). 0021-8693/$ – see front matter © 2007 Elsevier Inc. All rights reserved. doi:10.1016/j.jalgebra.2007.03.043
Y. Fan, B. Külshammer / Journal of Algebra 317 (2007) 250–259
251
proof made use of results in [4] and [5]. On the other hand, H. Meyer gives several examples of finite groups G such that Zp F G is not multiplicatively closed. In this paper, we will consider the F -subspace Zp A := A ∩ Zp F G of the center ZA of A, for a block A of a group algebra F G. As is well known, the dimension of Zp A coincides with the number of simple A-modules. Our main result will be a generalization of Meyer’s theorem to blocks. In Section 2 we show that, for a block A with an abelian defect group D (in a finite group with an arbitrary Sylow p-subgroup P ), the F -subspace Zp A is multiplicatively closed, i.e. a subalgebra of ZA. This generalization is not quite straightforward since Meyer’s proof uses a transfer argument which is not directly applicable to blocks. Our way around this problem is to use results by Watanabe [10], Fan [3] and Külshammer–Okuyama–Watanabe [7] instead. Our main result has applications to perfect isometries, as defined by M. Broué [1]. In Section 3 we show that, for blocks A and B with abelian defect groups in finite groups G and H , respectively, the isomorphism of F -algebras ZB → ZA induced by a perfect isometry between A and B maps Zp B onto Zp A. In particular, this applies to all situations where Broué’s abelian defect group conjecture is known to hold. Hence the F -algebra Zp A provides a new invariant of perfect isometries and hence of derived equivalences, at least for blocks with abelian defect groups. 2. Abelian defect groups In the following, we fix a prime number p and a p-modular system (K, O, F ), that is, O is a complete discrete valuation ring with field of fractions K of characteristic 0 and residue field F of characteristic p. The O-algebras we consider will always be free of finite rank as O-modules. It will be important, at a key step for our main result, to work over O, and not just over F , though the following first two results hold not only for O. We start with a somewhat technical result on abstract algebras whose relevance will become clear later. 2.1. Proposition. Let A be an O-algebra, and let L be a split local unitary subalgebra of the center ZA of A. Moreover, let e be an idempotent in A such that AeA = A, and suppose that C is a unitary subalgebra of eAe such that the multiplication map μC : L ⊗O C → eAe,
z ⊗ c → zc,
is an isomorphism of O-algebras. Then there exists a unitary subalgebra B of A such that C = eBe and such that the multiplication map μB : L ⊗O B → A,
z ⊗ b → zb,
is an isomorphism of O-algebras. r Proof. We write e = lr=1 m s=1 frs with pairwise orthogonal primitive idempotents frs in C where frs and fr s are conjugate in C if and only if r = r . Since L is local and since μC is an isomorphism, every frs is also primitive in eAe and thus in A. Moreover, frs and fr s are conjugate in A if and only if they are conjugate in eAe if and only if they are conjugate in C. Next, note that every primitive idempotent of A is conjugate to one in eAe since AeA = A; so we combine our decomposition of e with a similar decomposition of 1 − e in order to obtain
252
Y. Fan, B. Külshammer / Journal of Algebra 317 (2007) 250–259
r a decomposition 1 = lr=1 ns=1 frs into pairwise orthogonal primitive idempotents frs in A where frs and fr s are conjugate in A if and only if r = r , and nr mr for r = 1, . . . , l. We conclude that nr nr l
A=
frs Afr s ,
r,r =1 s=1 s =1
and eAe =
mr mr l r,r =1
s=1
frs Af
r s
,
C=
s =1
mr mr l r,r =1
s=1
frs Cfr s .
s =1
For r = 1, . . . , l and s = 1, . . . , nr , we can choose elements xrs ∈ frs Afr1 and yrs ∈ fr1 Afrs such that frs = xrs yrs and fr1 = yrs xrs ; here we may and will assume that xrs ∈ frs Cfr1 and yrs ∈ fr1 Cfrs whenever s mr . Then frs Afr s = xrs yrs Axr s yr s ⊆ xrs Ayr s ⊆ frs Afr s , and xrs Ayr s = xrs fr1 Afr 1 yr s ⊆ xrs (eAe)yr s ⊆ xrs Ayr s ; therefore, frs Afr s = xrs Ayr s = xrs (eAe)yr s for r, r = 1, . . . , l, s = 1, . . . , nr and s = 1, . . . , nr . And similarly, frs Cfr s = xrs Cyr s whenever s mr and s mr . This implies that, for r, r = 1, . . . , l, s = 1, . . . , nr , s = 1, . . . , nr , the maps xrs Ayr s → xr1 Ayr 1 ,
a → yrs axr s ,
xr1 Ayr 1 → xrs Ayr s ,
a → xrs ayr s ,
and
are mutually inverse isomorphisms of L-modules, which induce O-module isomorphisms xrs Cyr s ∼ = xr1 Cyr 1 . We claim that B :=
nr nr l
xrs Cyr s
r,r =1 s=1 s =1
is a unitary subalgebra of A satisfying the conclusion of the proposition. In fact, we have 1=
nr l r=1 s=1
frs =
nr l r=1 s=1
xrs fr1 yrs ∈
nr l
xrs Cyrs ⊆ B.
r=1 s=1
Moreover, we see that xrs Cyr s · xr s Cyr s = 0 unless r = r and s = s , and that xrs Cyr s · xr s Cyr s = xrs Cfr 1 Cyr s ⊆ xrs Cyr s ⊆ B
Y. Fan, B. Külshammer / Journal of Algebra 317 (2007) 250–259
253
for all possible r, r , r , r , s, s , s , s . Thus B is a unitary subalgebra of A containing C; hence C = eCe ⊆ eBe. On the other hand, we have exrs Cyr s e = 0 whenever s > mr or s > mr , and exrs Cyr s e = xrs Cyr s whenever s mr and s mr . This shows that eBe ⊆ C, and we have proved that eBe = C. At last, since the multiplication map μB is certainly a homomorphism of O-algebras, it is enough to show that the following restriction of μB is an O-isomorphism μB |L⊗O (xrs Cyr s ) : L ⊗O (xrs Cyr s ) → xrs (eAe)yr s ,
z ⊗ xrs cyr s → xrs zcyr s .
The hypothesis that the multiplication map μC is surjective implies that the above map is surjective; and it is injective because the following isomorphism implies that the O-ranks of both sides are equal: μC
L ⊗O (xrs Cyr s ) ∼ = L ⊗O (xr1 Cyr 1 ) ∼ = xr1 (eAe)yr 1 = fr1 Afr 1 ∼ = frs Afr s .
2
We can use Proposition 2.1 in order to prove the following structure theorem for certain blocks with abelian defect groups. 2.2. Theorem. Let G be a finite group, and let A be a block of the group algebra OG with maximal A-subpair (D, bD ). Suppose that D is abelian and that Q := CD (NG (D, bD )) ⊆ Z(G). Then there exists a unitary subalgebra B of A such that the multiplication map μ : OQ ⊗O B → A,
x ⊗ y → xy,
is an isomorphism of O-algebras. Proof. We fix an idempotent i in A such that iAi is a source algebra of A. Results of Fan [3] (which is proved without restrictions on the size of O) and Külshammer–Okuyama–Watanabe [7] imply that there exists a unitary subalgebra C of iAi such that the multiplication map OQ ⊗O C → iAi,
x ⊗ y → xy,
is an isomorphism of O-algebras. Since AiA = A, the result follows from Proposition 2.1.
2
In fact, the theorem above (for sufficiently large O) was already contained in a preliminary version of [7]; however, it was not incorporated into the final version. We also note that the O of A in O[G/Q], which is a block with algebra B above is clearly isomorphic to the image A defect group D/Q when O is large enough. In the following, we fix a finite group G and a block A = OGe of the group algebra OG, with block idempotent e and defect group D. We denote by Gp the set of p-regular elements in G, and by OGp the O-sublattice of OG spanned by Gp . In general, neither OGp nor Zp OG := ZOG ∩ OGp are multiplicatively closed (not even when the Sylow p-subgroups of G are abelian), as easy examples show. Note that the p-regular class sums form an O-basis of Zp OG. We set Zp A := A ∩ Zp OG = (Zp OG)e; the last equality follows from a result of Iizuka (cf. [6]).
254
Y. Fan, B. Külshammer / Journal of Algebra 317 (2007) 250–259
2.3. Corollary. In the situation of Theorem 2.2, we have Zp A ⊆ ZB. Proof. Let C be a p-regular conjugacy class of G, with class sum C + := c∈C c in OG. We need to show that C + e ∈ ZB. Since the multiplication map μ : OQ⊗O B → A is an isomorphism of O-algebras, we have decompositions A=
uB
and ZA = μ Z(OQ ⊗O B) = μ(OQ ⊗O ZB) = uZB.
u∈Q
u∈Q
First we assume that the group algebra O c for c ∈ C is split semisimple, i.e. is isomorphic to a direct product of copies of O; in particular, c is an O-linear combination of idempotents in O c . Then we can write C + e = nm=1 αm jm with idempotents jm ∈ A (not necessarily orthogonal) and coefficients αm ∈ O. Since A/J (A) ∼ = B/J (B), every idempotent of A is conjugate to one in B; so we may choose units um in A such that im := um jm u−1 m ∈ B, for m = 1, . . . , n. Then C+e =
n m=1
αm jm ≡
n
αm im =: b
mod [A, A] ,
m=1
where b ∈ B, and [A, A] is the O-sublattice of A spanned by all commutators xy −yx (x, y ∈ A). And note that [A, A] = μ [OQ ⊗O B, OQ ⊗O B] = μ OQ ⊗O [B, B] = u[B, B]. u∈Q
On the other hand, we may write C + e = u∈Q uzu with uniquely determined elements zu ∈ ZB. Then uzu − b = C + e − b ∈ [A, A] = u[B, B], u∈Q
u∈Q
and we conclude that zu ∈ [B, B] whenever 1 = u ∈ Q. Hence zu ∈ ZB ∩ [B, B] = 0 (since we are working in characteristic 0) whenever 1 = u ∈ Q, so we obtain C + e = z1 ∈ ZB. Next assume that O c for c ∈ C is not split. Since c is p-regular, there is a finite extension O is split semisimple; hence, in of O such that O c ⊗O A = := O A
u∈Q
⊗O B) = u(O
uB
u∈Q
⊗O B, we have C + e ∈ ZB := O (the arguments of the previous paragraph are in fact where B based on the structure A = u∈Q uB and independent of the two-sided indecomposability of A). So
= = ZB. uZB ∩ ZB 2 C + e ∈ ZA ∩ ZB u∈Q
Now we turn our attention to characteristic p. In the following, we denote the canonical maps O → F and OG → F G by x → x. Moreover, we denote by F Gp the F -subspace of the group
Y. Fan, B. Külshammer / Journal of Algebra 317 (2007) 250–259
255
algebra F G spanned by Gp . Then Zp F G := ZF G ∩ F Gp is the F -subspace of ZF G spanned by the p-regular class sums of G. Since A is a block of OG with defect group D, its image A is a block of F G with defect group D. We are interested in the F -subspace Zp A := A ∩ Zp F G = (Zp F G)e of ZA. Note that Zp F G = Zp OG and Zp A = Zp A. We call A p -closed if Zp A is multiplicatively closed, i.e. a subalgebra of ZA. Similarly, we call OG p -closed if Zp F G is a subalgebra of ZF G. By H. Meyer’s result [8], group algebras of groups with abelian Sylow p-subgroups are p -closed. We prove the following generalization. 2.4. Theorem. Blocks with abelian defect groups are p -closed. Proof. First we assume that O is large enough. Let A = OGe be a minimal counterexample, where e denotes the block idempotent of A. Let (D, bD ) be a maximal A-subpair, so that D is a defect group of A. Since D is abelian, we have D = Q × R where Q := CD (NG (D, bD )) and R := [D, NG (D, bD )]. Assume that z1 , z2 ∈ Zp A but z1 z2 ∈ / Zp A, and let z1 z2 =
zg g,
zg ∈ F.
g∈G
Then there is a p-singular element g ∈ G such that zg = 0. Writing g = su = us with a p-element u ∈ G and a p-regular element s ∈ CG (u), and applying the Brauer homomorphism Br u , we obtain Br u (z1 ), Br u (z2 ) ∈ Zp F CG (u)Br u (e), but Br u (z1 )Br u (z2 ) ∈ / Zp F CG (u)Br u (e). Since the defect groups of the blocks appearing in F CG (u)Br u (e) ¯ are still abelian, one of these blocks is still a counterexample to the theorem. By the minimality of the counterexample A = OGe, it must be the case that CG (u) = G. So u belongs to D and is centralized by NG (D, bD ), hence belongs to Q. We conclude that the p-factor of any g ∈ G such that zg = 0 belongs to Q. Let bQ be the unique block of OCG (Q) such that (Q, bQ ) (D, bD ). Then a result by Watanabe [10] implies that the map β : ZA → ZbQ ,
y → BrQ (y)1bQ ,
is an isomorphism of F -algebras; this isomorphism maps Zp A onto Zp bQ , so that bQ is also a counterexample to the theorem. Thus it must be the case that CG (Q) = G, in other words, Q is a central p-subgroup of G. Now, on one hand we may write z1 z2 =
u∈Q
us u =
u∈Q
us u e
256
Y. Fan, B. Külshammer / Journal of Algebra 317 (2007) 250–259
with s u ∈ Zp F G, hence s u e ∈ Zp A for u ∈ Q. On the other hand, by Theorem 2.2, we have a unitary subalgebra B of A such that A= uB, and ZA = uZB, u∈Q
u∈Q
where B denotes the image of B in A. Then, by Corollary 2.3, we have Zp A ⊆ ZB and therefore Zp A ⊆ ZB. In particular, ZB contains z1 , z2 and s u e for u ∈ Q. But then also z1 z2 ∈ ZB. Since ZA = u∈Q uZB we conclude that s u e = 0 whenever 1 = u ∈ Q. This shows that z1 z2 = s 1 e ∈ Zp A, and we have reached a contradiction. and extend A to If O is not large enough, then we can extend it to a large enough one O, ⊗O A. It is easy to see that = O A = Zp A. ZA ∩ Zp A decompose e as a sum of primitive central idempotents: e = e1 + · · · + en ; then e1 , . . . , en In A, t = Ae t for t = 1, . . . , n are blocks with defect are conjugate under a suitable Galois group, and A group D. For z1 , z2 ∈ Zp A, by the conclusion proved above, t , z1 z2 et ∈ Zp A
t = 1, . . . , n;
so z1 z2 =
n t=1
z1 z2 e t ∈
n
t = Zp A; Zp A
t=1
in conclusion, we get = Zp A. z1 z2 ∈ ZA ∩ Zp A
2
3. Perfect isometries In this section we assume that K and F are splitting fields for all the algebras we will consider, and we point out a connection between Theorem 2.4 and perfect isometries, as defined by M. Broué [1]. We start by recalling the definition of a perfect isometry and some of its consequences. Suppose that A = OGe and B = OHf are blocks of finite groups G and H , respectively; here e and f are the corresponding block idempotents. We denote the set of irreducible characters of G in A by Irr(A), the group of virtual characters of G in A by Z Irr(A), and the set of class functions G → K in A by CF(A, K). An isometry I : Z Irr(B) → Z Irr(A) (with respect to the canonical inner products) is called perfect if the virtual character μ of G × H defined by (Iβ)(g) · β h−1 (g ∈ G, h ∈ H ), μ(g, h) := β∈Irr(B)
satisfies the following two conditions, for g ∈ G and h ∈ H :
Y. Fan, B. Külshammer / Journal of Algebra 317 (2007) 250–259
257
(Int) μ(g, h)/|CG (g)| ∈ O and μ(g, h)/|CH (h)| ∈ O. (Sep) If μ(g, h) = 0 then g and h are either both p-regular or both p-singular. Since I is an isometry, each β ∈ Irr(B) defines a sign β ∈ {1, −1} and an irreducible character αβ ∈ Irr(A) such that Iβ = β αβ . The map β → αβ is a defect-preserving bijection between Irr(B) and Irr(A), and I extends to a K-linear bijection, also denoted by I , between CF(B, K) and CF(A, K). Note that (I ψ)(g) = |H |−1
μ(g, h)ψ(h)
ψ ∈ CF(B, K), g ∈ G .
h∈H
Recall that the group algebra KG is a symmetric K-algebra with respect to the canonical bilinear form(· | ·) mapping a pair of elements x, y ∈ KG to the coefficient z1 ∈ K in the product xy = z = g∈G zg g. The restriction of (· | ·) turns KA := K ⊗O A into a symmetric K-algebra. Hence the vector spaces ZKA and CF(A, K) are isomorphic via the map associating to each class function φ ∈ CF(A, K) the element φ ◦ ∈ ZKA satisfying (φ ◦ |a) = φ(a) for a ∈ KA. Explicitly, we have φ g −1 g. φ◦ = g∈G
In a similar way, ZKB and CF(B, K) are in canonical K-linear bijection. Thus the K-linear bijection I : CF(B, K) → CF(A, K) yields a K-linear bijection I ◦ : ZKB → ZKA such that I ◦ ψ ◦ = (I ψ)◦ for ψ ∈ CF(B, K). For β ∈ Irr(B), let fβ = (β(1)/|H |) h∈H β(h−1 )h = (β(1)/|H |)β ◦ denote the primitive idempotent in ZKB corresponding to β. Similarly, for α ∈ Irr(A), let −1 α g g = α(1)/|G| α ◦ eα = α(1)/|G| g∈G
denote the primitive idempotent of ZKA corresponding to α. Then I ◦ fβ = β(1)/|H | I ◦ β ◦ = β(1)/|H | β αβ◦ = β β(1)/|H | |G|/αβ (1) eαβ for β ∈ Irr(B). Since f =
β∈Irr(B) fβ
I ◦f =
we conclude β β(1)/|H | |G|/αβ (1) eαβ .
β∈Irr(B)
In general, I ◦ is K-linear but not necessarily a homomorphism of K-algebras. It is easy to see that |H |−1 I ◦z = μ g −1 , h zh−1 g for z = zh h ∈ ZKB. g∈G
h∈H
h∈H
Thus the condition (Int) implies that I ◦ restricts to an O-linear map ZB → ZA which we will also denote by I ◦ .
258
Y. Fan, B. Külshammer / Journal of Algebra 317 (2007) 250–259
3.1. Lemma. In the situation above, we have I ◦ (Zp B) = Zp A. / Hp , and I ◦ z = g∈G yg g with Proof. Let z = h∈H zh h ∈ Zp B. Then zh = 0 whenever h ∈ yg = |H |−1 h∈H μ(g −1 , h)zh−1 for g ∈ G. If yg = 0 then 0 = μ(g −1 , h)zh−1 for some h ∈ H , so h ∈ Hp since zh−1 = 0. But now (Sep) implies that g ∈ Gp , and we have proved that I ◦ (Zp B) ⊆ Zp A. Thus the lemma follows by symmetry. 2 ◦
The O-linear bijection I ◦ : ZB → ZA induces an F -linear bijection I : ZB → ZA such that ◦ I (Zp B) = Zp A. In general, I is not a homomorphism of F -algebras, as easy examples show. On the other hand, the perfect isometry I between A and B induces an isomorphism of K-algebras ◦
ι : ZKB → ZKA in such a way that ι(fβ ) = eαβ for β ∈ Irr(B). The maps ι and I ◦ are related by the formula (see [1]) ι(z) = I ◦ zR ◦ (e) for z ∈ ZKB; here R = I −1 denotes the perfect isometry inverse to I . Hence ι restricts to an isomorphism of O-algebras ZB → ZA also denoted by ι. Note that e ∈ Zp A by a result of Osima (cf. [6]). Thus R ◦ (e) ∈ Zp B by Lemma 3.1. We denote by ι : ZB → ZA the isomorphism of F -algebras ◦ ◦ induced by ι : ZB → ZA. Then ι(z) = I (zR (e)) for z ∈ ZB. 3.2. Proposition. Let I be a perfect isometry between the blocks A = OGe and B = OHf , and suppose that B is p -closed. Then A is also p -closed, and the isomorphism of F -algebras ι : ZB → ZA defined by I satisfies ι(Zp B) = Zp A. ◦
◦
Proof. Let z ∈ Zp B. Since R (e) ∈ Zp B and since B is p -closed, we conclude that zR (e) ∈ ◦ ◦ Zp B. Thus Lemma 3.1 implies that ι(z) = I (zR (e)) ∈ Zp A. So we have proved that ι(Zp B) ⊆ Zp A. Since dim Zp B = dim Zp A, we obtain ι(Zp B) = Zp A, and the second assertion is proved. Since Zp A is the image of the subalgebra Zp B under the isomorphism of F -algebras ι, it is certainly a subalgebra of ZA, and the result follows. 2 We obtain the following consequence. 3.3. Corollary. Let I be a perfect isometry between blocks A = OGe and B = OHf of finite groups G and H , respectively, and suppose that the defect groups of B are abelian. Then A and B are p -closed, and the isomorphism of F -algebras ι : ZB → ZA defined by I maps Zp B onto Zp A. Proof. By Theorem 2.4, the block B is p -closed. Hence the result follows from Proposition 3.2. 2 The result above implies that the F -algebra Zp A is an invariant of perfect isometries and hence an invariant of derived equivalences, at least for blocks with abelian defect groups. (We
Y. Fan, B. Külshammer / Journal of Algebra 317 (2007) 250–259
259
recall that the dimension of Zp A is just the number of simple A-modules.) In the situation of Corollary 3.3, it is not true, however, in general, that ι(Zp B) = Zp A, as elementary examples show. Of course, one would expect that, in the situation of Corollary 3.3, the defect groups of A are also abelian. (This would follow, for example, from Brauer’s height zero conjecture.) One would perhaps even expect that A and B have isomorphic defect groups. The result above motivates the following problem. 3.4. Question. Let ι : ZB → ZA be the isomorphism of F -algebras defined by a perfect isometry I between blocks A = OGe and B = OHf of finite groups G and H , respectively. Is ι(Zp B) = Zp A? J. Murray [9] has proved that blocks of finite symmetric and alternating groups are p -closed. Thus, by Proposition 3.2, Question 3.4 has a positive answer whenever H is a symmetric or alternating group. (We also note that M. Enguehard has proved that p-blocks of the same weight in finite symmetric groups are always perfectly isometric [2].) One may view Question 3.4 as a problem concerning the p-sections of 1 in G and H , respectively. Thus one may ask whether similar properties hold for other p-sections as well, at least in the presence of an isotypy (cf. [1]). However, this does not seem to be the case, as easy examples show, not even for blocks with abelian defect groups in the situation of Broué’s abelian defect group conjecture. Acknowledgments The results in this paper were obtained during visits of the first author in Jena and of the second author in Wuhan. Both visits were made possible by a Sino-German exchange program dedicated to the theory of finite groups and their representations, funded by the NSFC and the DFG. Both authors gratefully acknowledge this support. The second author also profited from several discussions with G. Cliff, on the occasion of a visit to the University of Alberta in Edmonton and on the occasion of a visit by G. Cliff in Jena. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10]
M. Broué, Isométries parfaites, types de blocs, catégories dérivées, Astérisque 181–182 (1990) 61–92. M. Enguehard, Isométries parfaites entre blocs de groupes symétriques, Astérisque 181–182 (1990) 157–171. Y. Fan, Relative local control and the block source algebras, Sci. China Ser. A 40 (1997) 785–798. B. Külshammer, Bemerkungen über die Gruppenalgebra als symmetrische Algebra II, J. Algebra 75 (1982) 59–69. B. Külshammer, Group-theoretical descriptions of ring-theoretical invariants of group algebras, in: Progr. Math., vol. 95, 1991, pp. 425–442. B. Külshammer, Central idempotents in p-adic group rings, J. Aust. Math. Soc. Ser. A 56 (1994) 278–289. B. Külshammer, T. Okuyama, A. Watanabe, A lifting theorem with applications to blocks and source algebras, J. Algebra 232 (2000) 299–309. H. Meyer, On a subalgebra of the centre of a group ring, J. Algebra 295 (2006) 293–302. J. Murray, Squares in the centre of the group algebra of a symmetric group, Bull. London Math. Soc. 34 (2002) 155–164. A. Watanabe, Note on a p-block of a finite group with abelian defect group, Osaka J. Math. 26 (1989) 829–836.